GST 101 Introduction to Geospatial Technology � �Unit 8 - Introduction to Remote Sensing and Aerial Imagery �Module 8.3 – Viewing and Basic Analysis Methods for Remote Sensing Data ����� �

Ann Johnson

Associate Director

ann@baremt.com

​

Based upon work supported by the National Science Foundation under Grants DUE 1304591, DUE 164409, DUE 1700496, DUE 1937177, Due 1938717 DUE 1937237, 2030206 and 2015927. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation.

Module 8.2 Focused on Finding and Downloading an Image File of Landsat 8 Data

• This Module 8.3 will look at downloading data and how to use it to visualize an area of interest and carry out some basic analysis methods
• Different software packages can be used, but this example will use ArcGIS Pro from Esri
• Before downloading the image, a workspace should be set up on the computer or on the Server where it will be stored and accessed
• Once downloaded it should be moved into the workspace and unzipped

​

Image Analysis: Art Versus Science

• Image Analysis: Act of examining images for the purpose of identifying and measuring objects and phenomena and judging their significance
• Image interpretation is not an exact science
• Interpretations tend to be probabilistic not exact
• Successful interpretation depends on user’s
• Systematic and disciplined approach using concepts of remote sensing
• Knowledge of software and other applications used in the examination
• Knowledge of other factors (geography, geology, land use, and many other disciplines) needed to effectively use remote sensing imagery data

​

What Does Remote Sensing Imagery Data Looks Like ��

Downloaded original data unzipped twice

Gray scale image from one band using brightness from Digital Numbers

Landsat imagery band data from multiple sensors

• Remote Sensed Imagery is downloaded as a zipped file containing multiple files as shown on the left
• The middle image is an illustration of the data files for imagery bands
• Each band can be displayed using a “gray scale” with brightness based on Digital Number values for pixels from lowest values (Black) and highest values (White) based on energy detected by a sensor for each pixel

Top of Atmosphere

Reflected

Solar Irradiance Is the amount of energy provided by the Sun (Watts/meter² * srad * μm)

​

Reflected energy varies due to effects on irradiance from Earth/Sun geometry (orbital distance and Tilt)

​

Landsat Collection 2 data adjusts for these effects and provide consistent reflectance values for locations collected on different dates

Brightness Energy Levels (Digital Numbers) for a Band

Landsat 8, Band 4 scene

Band 4 Brightness of Buildings

Brightness (Digital Number) of Pixel – larger number brighter (more energy) for that band

Pixels – 30 x 30 m

Correction of Landsat Imagery Data � Reflectance Collection Level 1 to Collection 2 Level 2�

Landsat Level 1, Top of Atmosphere

Collection 2 Level 2 surface reflectance

Surface temperature

Combining Bands to Create Composite Images

Imagery Data Can Be Used to Visualize Wavelengths that Cannot be Seen by the Human Eye

• Remote sensing software can use the grayscale Digital Numbers of a band to display it as one of three primary colors using color guns on computer monitors with saturation based on its brightness values
• Red, Green and Blue (RGB)
• This includes displaying wavelength bands (infrared, near infrared, etc.) not seen by most humans by assigning one of the three colors to these bands
• Up to three bands cam be visualized to create a composite image using the values for brightness and the three primary color guns
• One band to be used for Red
• One band to be used for Green
• One Band to be used for Blue

Composite Images

Gray Scale Brightness values (DN) from three Bands are combined and displayed by assigning each value to either the red, green, or blue color gun on a computer monitor creating an image:

Natural, False and Pseudo

Esri.com help

Landsat 8 Onion Skin

• This NASA video shows what you can visualize using different bands as composite images using Landsat 8 bands:

http://svs.gsfc.nasa.gov/cgi-bin/details.cgi?aid=11491

Each Landsat Mission includes sensors that collect data in�specific bands and users should be sure to use the correct band for the Mission

• Band 1: 450 - 520 nm (Red)
• Band 2: 520 – 600 nm (Green)
• Band 3: 630 – 690 nm (Blue)
• Band 4: 760 - 900 nm (Near infrared)
• Band 5: 1550 – 1750 nm (Mid-Infrared)
• Band 6: 10400 - 12500 nm (Thermal infrared)
• Band 7: 2080 - 2350 nm (Mid-infrared)

6 7 5 4

Our Eyes

Landsat 7

Spectral Resolution of Different Landsat Missions

Watch Spectral Resolution Concept Module at https://www.youtube.com/watch?v=3xHjiXloif4

Different Landsat 5 and 7 and 8 Missions Use Different Band Numbers to Create Composite

Band numbers for Landsat 5 and 7 are different than for Landsat 8 – be sure you are using the correct bands

Landsat 8 - Composites�� ` Natural or True Color� Bands 4, 3, 2� � False Color� Band 5, 4, 3��� Pseudo Color� Bands 7, 5, 3��

Comparison of Landsat 7 and 8 and Sentinel-2 band numbers� Note: This graphic shows good atmospheric widows using gray shading

Landsat 8 band numbers

Landsat 7 band numbers

Sentinel 2 band numbers

Remote Sensing and Image Analysis Workflow Questions

Landsat Explorer – A Quick Look at Your Area of Interest

• Landsat Image files are large and can take time and resources to download
• Before determining what imagery you may need, you can explore your area of interest and create composite images using your browser and LandsatExplorer from this link: https://livingatlas2.arcgis.com/landsatexplorer/
• For Example: I want to study urbanization in Northern Idaho near Rathdrum
• Using LandsatExplorer and entering Rathdrum, ID USA and searching

LandsatExplorer Tools -

Composited Bands with description and what bands are used

• Agriculture : Highlights agriculture in bright green; Bands 6, 5, 2
• Natural Color : Sharpened with 15m panchromatic band; Bands 4, 3, 2 +8
• Color Infrared : Healthy vegetation is bright red; Bands 5, 4 ,3 
• SWIR (Short Wave Infrared) : Highlights rock formations; Bands 7, 6, 4
• Geology : Highlights geologic features; Bands 7, 6, 2
• Bathymetric : Highlights underwater features; Bands 4, 3, 1
• Panchromatic : Panchromatic images at 15m; Band 8
• Vegetation Index : Normalized Difference Vegetation Index(NDVI); (Band 5 - Band 4)/(Band 5 + Band 4)
• Moisture Index : Normalized Difference Moisture Index (NDMI); (Band 5 - Band 6)/(Band 5 + Band 6)
• SAVI : Soil Adjusted Veg. Index); Offset + Scale*(1.5*(Band 5 - Band 4)/(Band 5 + Band 4 + 0.5))
• Water Index : Offset + Scale*(Band 3 - Band 6)/(Band 3 + Band 6)
• Burn Index : Offset + Scale*(Band 5 - Band 7)/(Band 5 + Band 7)
• Urban Index : Offset + Scale*(Band 5 - Band 6)/(Band 5 + Band 6

​

Landsat Viewer

• View mosaiced images using a browser
• Useful for initial investigation
• Also as a source to download data

Adding the Bands to ArcGIS Pro and Creating a Natural Color Image

• Landsat 8:

Using bands 4, 3, 2

Landsat 8: Using Band 7, 5, 3 for Pseudo Color

• Highlight's water, vegetation, bare soil and urban areas
• Gray shaded streak is shadow from a cloud or possibly a contrail from an aircraft
• Note: that any 3 bands can be combined, and this may be useful for users that have color deficiencies

Identifying and Classifying Features

• Features can be visually investigated using composite images made of the different band combination
• Visualizations of patterns can be created by using “band algebra” equations for specific applications such as Normalized Difference Vegetation Index (NDVI) using Near Infra Red and Red bands to identify greenness of vegetation

Tool using Composite Band to create NDVI Using NIR and Red Band DNs in the NDVI equation:

NDVI = (NIR + Red)/(NIR – Red)

Information about Landsat and other Missions

• Video from NASA and Landsat Wavelengths

https://www.youtube.com/watch?v=YP0et8l_bvY

• See USGS site for current news and resources about Landsat Collections
• https://www.usgs.gov/core-science-systems/nli/landsat

Spectral Signature Graphs – Spectral Profiles in ArcGIS Pro

• Spectral signature graphs can help identify the feature compositions (soil, vegetation, water, etc.) of individual pixels by graphing spectral reflectance values of multiple bands for the pixel
• The information can be used to help in different classification techniques to identify or validate classification outcomes

A combined graph that compares different Land Use classification spectral signatures.

Graphs from two time-frames can be used to distinguish healthy versus stressed vegetation as shown in Figure 3 above

​

Example of a Spectral Profile in ArcGIS Pro

LandsatExlorer using “i” tool can also be used to create a Spectral Signature Profile Of individual Pixels to Help Identify Features

Spectral signatures are graphs of values for all wavelengths of one pixel - graph of a pixel identified as urban from Esri LandsatExplorer for image from Rathdrum, ID

​

Classification Using Software Tools

• Band data from remote sensed imagery can be used to identify features (water, vegetation, road, building, trees, etc.) and create land cover images for a study area
• Classification can be based on per pixel, sub pixel and object-based methods using geospatial technology
• Two methods for image classification are Unsupervised and Supervised Classification
• Unsupervised classification allows the software to autonomously group like pixels or objects by their band data into a specified number of groups. The user then uses a Classification Schema to manually color groups into feature types, and merge groups together creating a land cover image
• Supervised classification requires the user to initially have the computer create a segmented image of like pixels or objects. The user then creates a training set of values for different land cover features from the imagery. The software then uses the training set to group the data into the Classification Schema of land cover classes. This process can be repeated and manually edited to create a final land cover image.

Pixel Based or Object-Based Classification

• Pixels based image analysis groups pixels based on their pixel spectral values
• All pixels with the same spectral value are color coded the same random color based on how many classes the user has specified. User then adjusts colors for different feature categories (water, urban, soil, etc.)
• Objects based image analysis groups pixels together based on the similarity of their nearby (using algorithms such as spatial autocorrelation) pixel values so that contiguous groups (even if different spectral values) are coded the same color

Mixed Pixel – Spatial and Spectral Resolution and Classification of Imagery

Addressing the Mixed Pixel problem using spatial autocorrelation and Artificial Intelligence combined with field verification and other techniques may be useful for sub-pixel classification methods. Tools based on these techniques are included in ArcGIS Pro help provide a “sub-pixel” value for a mixed pixel.

Tools in ArcGIS Pro:

​

Nearest Neighbor

​

Bilinear

4 cells

​

​

Cubic

16 cells

​

Majority

Most popular

value of 4 cells

Unsupervised Classification in ArcGIS Pro

Supervised Classification

Steps Within ArcGIS Pro Wizard

​

• Using a composite of bands for an image, user identifies type of Classification – in this case Supervised and either pixel or object based and specifies a Classification Schema (default is the National Land Cover data set and then perform an initial segmentation
• The next step is to create a training sample using the training sample manager for the segments from the first step. This is a process that links the band data of segments to the type of feature in the classification schema using one of the tools in the Wizard
• Then use the tool to select features in the image that match the feature type in the schema
• The wizard asks for the Classifier method to be used on the training sample and the number of samples
• The run command is used which outputs a classified image. The output can then use the Wizard to merge classes and subclasses into merged classes
• Finally, the Wizard “reclassifier” can be used to manually edit the output by reclassify object or regions
• See the URL from Esri on the right for specific steps

https://www.esri.com/videos/watch?videoid=Cg_JOMJazh8&title=introduction-image-classification-in-arcgis-pro

Spectral Profile – Signature Graphs Can Help Classify Images

Water

Urban – Built Environment

USGS Site of For High Resolution Spectral Library

• Spectral Library Version 7
• https://crustal.usgs.gov/speclab/QueryAll07a.php
• Provide data for different substances using lab, field and imaging spectrometers
• Can help identify substance from remote sensing data

Comparison of spectra for alunite from four sensors with different spectral resolutions. Figure 6 from USGS Circular 1413.

Remote Sensing

• This Unit 8 provide a brief overview of some of the concepts important for use of remote sensing imagery
• A full course in remote sensing is highly recommended as there are many more uses and techniques available for integrating remote sensing into GIS and geospatial projects
• See the GeoTech Center Introduction to Remote Sensing Model Course

See GeoTech Center website (https://geotechcenter.org) �for additional Model Courses and other curriculum resources. ������Note: some content is a derivative of other authors���

Ann Johnson

Associate Director

ann@baremt.com

​

3-17-2021 V10